Climate Data Hack

What is the problem statement?

SD04 How can we support communities and organisations in ‘data poor’ environments such as in the Sahel and parts of Southern Africa to know when and where to get ready for a disaster? Many places where disasters occur often lack vulnerability and exposure data about communities at risk. Hydro-meteorological data can also be sometimes lacking. These data points are critical to creating triggers protocols which hydro-meteorological services and communities agree on and early action protocols that are carried out once a forecast reaches a certain threshold. This data can help prompt early humanitarian action, which can help save lives and reduce the impact of the disaster. What approach can be taken to collect data to feed into impact-based forecasts from data poor locations (with a focus on vulnerability and exposure data) in the Sahel and southern parts of Africa? -> Could we crowdsource data? How can volunteers act as agents to feed into impact-based forecasts? What data could local volunteers collect within a country to feed into impact-based forecasts? What would they collect? How would they collect and report it? How would it be visualised for decision makers to respond? What other datasets are useful for identifying exposure and vulnerability? E.g., How can OpenStreetMap data be used? This is useful for flood risk (mapping roads, houses and buildings) but might have less clear use for drought risk. What attributes might be useful to know (public or internal) for different disasters? What open data sets are available that can be used to support this. Sustainable Development: How can we support communities and organisations to know when to better prepare for emergencies by the prediction of a climate hazard? Sustainable Development 2: Support the most vulnerable communities Paul Knight and Caroline Zastiral (British Red Cross)

Our Ideas and Resources

Description/Idea Technique/Tech Data Comments
Using precipitation/temperature and drought indexes to better forecast when droughts and floods might occur. LSTM for time series forecasting: https://alexmoltzau.medium.com/predicting-floods-drought-with-ai-d959c26d4b65 Accessing historic Rainfall data from CHIRPS: https://github.com/planet-os/notebooks/blob/master/api-examples/CHIRPS_example.ipynb https://arbolmarket.medium.com/working-with-the-chirps-dataset-ed3f48545264 And CAMELS: https://pypi.org/project/camels/ https://earthdata.nasa.gov/earth-observation-data/near-real-time/hazards-and-disasters/drought Precipitation Data https://ral.ucar.edu/solutions/products/camels https://www.chc.ucsb.edu/data/chirps Drought Monitoring Info https://www.un-spider.org/links-and-resources/data-sources/daotm-drought
Using satellite imagery to define how built up an area is and the road network to get there and deliver aid. CNN on satellite images: https://towardsdatascience.com/loading-data-from-openstreetmap-with-python-and-the-overpass-api-513882a27fd0 https://developers.arcgis.com/python/sample-notebooks/automatic-road-extraction-using-deep-learning/
Understand the size/wealth of the population affected. Dataset discovery rather than anything more https://developers.arcgis.com/python/sample-notebooks/population-exploration-dashboard/ https://stats.oecd.org https://geoagro.icarda.org/en/cms/category/maps/16/regional
Get some geographical visualisations working. Streamlit – seems like a good option as an interactive app tooling from python https://medium.com/spatial-data-science/build-and-deploy-full-web-applications-quickly-with-python-e196771f48df https://towardsdatascience.com/build-a-multi-layer-map-using-streamlit-2b4d44eb28f3 https://developers.arcgis.com/python/sample-notebooks/building-a-change-detection-app-using-jupyter-dashboard/

Extra resources:

Datasets:

Overall Solution – Playback

Additional Info

For further info - please contact any of the team